Slashdot Mirror
Rearranging Pixels For Performance
tepes writes "From bottomquark, A new method of sub-pixel rendering could make monitors cheaper to produce. ClairVoyante Laboratories developed the PenTile Matrix, which uses five sub-pixels instead of the typical three, to take advantage of the fact that the human eye is more sensitive to blue colors."
The human eye is least sensitive to blue. It it most sensitive to green, followed by red then blue. R
The magic recipe will eventually be hit upon - monitors and video screens of very high quality will be produced cheaply, and people will begin putting them everywhere.
There are many uses for ubiquitous screen technology. But the more video we see and watch during a day the closer we get to a certain question. Will the video representation of reality become more comfortable for people than the real thing?
Many people already see more TV than they do real world outdoor imagery during a day. What happens when we all do? At least one issue to consider is that the cultural norms for the appearance of a healthy, sexually appealing human being will have even more to do with TV than they do today.
Goat sex free since 2001
I can see this is probably going to be like mp3 compression, where people often say "I can tell the difference between that and the original". So, someone will have to develop something silly like "monitor drivers with more blue!"
-- Dan
As with all slashdot posts, the posting is inaccurate.
The human eye is *least* sensitive to blue... that's what this thing is about, sort of.
It's also not a new method of sub-pixel rendering.. it's a new method of sub-pixel layout.
The theory is that in a conventional LCD, there is too much blue.. it's wasted space, resources, etc.
This thing both changes the color proportion, and the way the thing is wired up. adjacent subpixels of the same color are driven by the same driver.
Won't this make fonts look even more fuzzy and have more "jaggies"? Why aren't there any 3072x2304 monochrome laptops available? Doesn't anyone else think it's a good goal to have dynamic paper-quality images rather than pixels we are able to casually count?
It's not so much that we are less sensitive to blue.. that would seem to suggest you need MORE blue..
It's that we percieve blue at a lower resolution than the other colors, due to wavelength. (Ever seen blueblocker sunglasses? There is truth to that). So there is no reason to have the same resolution of R, G, and B, the extra resolution on B is wasted (not brightness, just resolution)
Hmm, hard to find a definitive source. But, some support for that assertion is here ("10Eh : 320x200 64k-colour (5:6:5)", "111h : 640x480 64k-colour (5:6:5)", ...) and here ("16 bit color depth is supported through several different bit arrangements, including 5-5-5 and 5-6-5.").
Briefly...
It is a really well written desription, it is a shame Design Engineering didn't have an writer that could understand it.
In the printing industry there has been a big trumpetting of a new dot layout
called Hexachrome. This takes the concept of the human eye's propensity
for blurring colors together and adjusts the traditional 3 dot 3 color priniting
layout to 6 dots with a red dot, a blue dot, and 4 different types of yellow/green.
Personally, this looks like a migration from paper to the computer
industry of this same technique which affords more vibrant colors and
cleaner details.
--Ks9
No it's not. Do a google search on CLEARTYPE to find pages describing the technique that the Apple ][ was using (Woz == God) and that M$ tried to patent twenty years after.
/. think about a pattern that would represent this even better scheme?
But that's funny you mention that, because CLEARTYPE using the contiguous alternation of red green and blue on the screen WON'T work with this new thing. And because MS had to go to the trouble of many thousand hours of user testing to optimize CLEARTYPE, expect them to anti-market this type of technology (as they clearly did with BlueTooth in XP.)
One more point: Is it me or the figure showing the pattern shows the blue as covering pretty much as much as the other components?
My concern (if the figure is not right), does that work for people who are colorblind or are more sensitive to blue than the average user population?
Finally just a question: why did they stop here, as green is perceived twice as much as red and red is perceived twice as much as blue (roughtly.) Can anyone on
PPA, the girl next door.
-- I feel better now. Thanks for asking.
I think you completely misunderstood what I meant.
Why do you think blue isn't picked up by your eyes as well? DUH! WAVELENGTH!
Blue does not focus at the same distance as the other colors, by enough of a margin to make excess blue make images seem fuzzy.
Sheesh.
Quote from the article: "The human eye, however, perceives blue at a much lower resolution than red and green."
They're saying blue is the *worst*. Also, I think sensitvity is different than resolution.
Rearranging the color pixels for color stealing is a reasonable idea, and making blue a little bigger is a nice tweak. Is it worth it? That's difficult to say. Subpixel rendering using color stealing on current LCDs actually does roughly put the extra resolution where you want it for high resolution text--vertical lines are the problem in small text, no horizontal lines.
Okay, I agree that this technology is cool, but I think I would still opt for a traditional LCD display. I'm red-green colorblind, so I am most sensitive to blue, rather than red or green as this display assumes.
I'm surprised that nobody else has posted about colorblindness yet-- I was under the impression that more of us engineering types were affected!
For example, in the case of a 15' 1600 x 1200 UXGA panel
Good God man! This would look great behind a 12" stonehenge...
I demand a million helicopters and a DOLLAR!
That's why in 16-bit color, it's 5 bits for red, 6 bits for green, and 5 bits for blue.
More precisely, the agreed upon values for the relative sensitivity of R,G,B are 0.3, 0.59, 0.11. To know how the 16 bits should be allocated, you find c such that:
(c+log2(0.3)) + (c+log2(0.59)) + (c+log2(0.11)) = 16
c=7.23, so the optimal split would be:
5.49 + 6.47 + 4.04 = 16 bits
It's not just you, that's how human vision works in general, and it's exactly the effect this new layout exploits. There are fewer blue pixels (which are therefore further apart) because our spatial perception of blue is less exact than of red/green. In effect, since we can't see the higher resolution of blue, why provide it? Having the same number of blue pixels as red or green is wasting a lot of pixels -- due to the "bluriness" in which we perceive blue, there's no point in having that many blue pixels -- we can't tell the difference between that and only having half as many blue pixels.
"Convictions are more dangerous enemies of truth than lies."
The human eye has two different types of photoreceptors called cones and rods. Cones
come in three variations for color vision.
But they don't perform too well in low light
conditions.
Rods can only perceive green or yellow light
and are much more sensitive.
That's why your color vision is reduced at
night and it's why it's so hard to see blue
stuff in low color conditions.
Apparently the column drivers on an LCD cost more than the row drivers. I have no idea why, but I will accept that.
It may have to do with the order that the row and column are printed in. If the rows are printed first, and the LCD has defects at that point, you only lose the time and money that went into the first few steps. Then if the Columns are printed later, and the LCD has defects, you lose that much more time and process cost. Thus minimizing the complexity and reducing the chance of defects in later printing stages is a wise move, from an economical standpoint.
At the fab I briefly contracted for, no one cared if you dropped the US$10 raw wafers, but people flipped out if you dropped the US$300 processed wafers that came out of the implanter 8 hours later...
Won't this screw up Cleartype? At least until they have an option to support this particular sub-pixel organization. Does Cleartype support multiple sub-pixel orderings right now? Although this seems like it would be a bit more complex, since the ordering changes not just on the x axis, but is differently laid out on the y axis as well.
Sounds like an interesting problem. I wonder how much information modern lcd displays give the cpu about their sub-pixel layout.
Which brings me to another question -- I wonder if anyone has looked into designing an image format which contained extra data to allow sub-pixel display layout of the image? Or whether there are any image display programs that take advantage of sub-pixel layout when scaling. Or further, hardware scaling routines on laptops (for when you're at lower resolution) that use it. (On the other hand, images are probably more color sensitive than text, so this might not work nearly as well).
Well, random thoughts.
-Puk
In 1953, the National Television System Committee was given the task of implementing a 3:1 compression scheme that could be decoded with a couple of extra vacuum tubes. All the frequencies had been assigned to work with simple black and white televisions. Furthermore, the new color signals had to work with unmodified black and white television sets, a large existing installed base.
The engineering that they did for this was completely brilliant and used the same kind of reasoning about the perceptual properties of the human eye that this product does. It got the job done. And we've been suffering the consequences for fifty years. Anyone who has ever done serious animation for video knows about chroma crawl, notch and comb filters, antialiasing along several different color axes at once, yada yada yada.
With LCD's and decent CRT's, we've been able to get away from most of that, unless you really need to put your signal on a television set. And so now, to save a few bucks on something that is going quickly down in price anyway, we're going to be hobbled for another fifty years because of some "clever" idea? This is progress?
Checking out the linked page, there are explanatory graphics midway down, but they're simply wrong. They show an enlarged letter A, (black on white) then show how that letter is formed on 'stripe' CRTs vs their tile system. The problem is that they have it reversed. They show the color phosphor dots on the black areas and the white areas are still white. The more fundamental difference here is that CRTs are additive, while LCD displays are subtractive, but they don't even go into that.
Worse, they base their assumptions of superiority on the misconception that striped CRT monitors have one trio of RGB stripes for each pixel. They don't even address the triangular RGB phosphor pattern that non-trinitron CRTs use.
In a nutshell, it sounds like a neat idea, but it's no panacea, and looks like it'll have many of the same edge-color problems that current CRTs do (Trinitron and non), only they'll be more obvious on 45deg angles of red and green surfaces, rather than 90deg angles. Take a look at the tile pattern, and see how the pattern does still have stripes, only they're rotated 45degrees right for green and 45 degrees left for red. I imagine a field of 100% blue will, on close inspection, be a thousand little points of light, since each one is surrounded by dark space that takes up 70% of the screen.
Of course, the proof is in the pudding. I wonder when they'll have samples at tradeshows.
Kevin Fox
Go ahead and ignore all but the first paragraph of my post. I was stupid.
The part about the 'A' graphic still stands though.
Kevin Fox
What makes column drivers more expensive is that they need to switch with a much higher freqency than the row drivers.
Microsoft has a bullshit patent on sub pixel rendering of text on LCD screens. I'll bet going to another pixel layout circumvents it and lets other companies write sub pixel rendering drivers without getting sued by big evil.
God, those patents are dumb. Of course you could find prior art, like the way every Apple II programmer drew fonts. I played with sub pixel rendering on my first color laptop just because it was such an obvious thing to try and it used to be my job to play with graphics...
Oh well enough whining. I'm a hypocrite, anyway. If my company puts my name on any software patents I won't complain about it, I'll be too busy wondering if that means that I'm getting a raise.
Rocky J. Squirrel
I note the date on the ClairVoyante web page is 1999. They've had this for a while, and still nothing's coming of it....
i thought its green..
Of the dark hues (RGB), green appears brightest to the eye.
But this article refers to resolution, detail - based on the concentration of blue cones on the retina. A person would have more difficulty reading tiny glowing blue text than green... so there's no point in providing that extra detail.
You probably experienced presbyopia, which is the gradual loss of flexibility in your lenses, thus making near items harder to focus on. This is a normal process that occurs with age.
Admittedly, looking at a monitor all day long can cause eyestrain, especially if you tend toward hyperopia (farsightedness) to begin with. That's my problem: While I can see all right at distances of a couple feet from my face and beyond, my eye muscles have to strain constantly, even when I'm focused at infinity, and working on a computer screen all day every day makes the strain get really bad if I'm not wearing my glasses.
"Biped! Good cranial development. Evidently considerable human ancestry."
No it's not. Do a google search on CLEARTYPE to find pages describing the technique that the Apple ][ was using (Woz == God) and that M$ tried to patent twenty years after.
Do another search, and you'll find the pages where Steve Gibson *retracted* that statement. The Apple II didn't have subpixel rendering. It simply used its pixel generator to create colors on top of a black & white NTSC signal by having a high enough resolution that the bandwidth of the signal crossed over into the area reserved for chroma data.
There's a big difference there.
Try reading the actual ClearType papers too -- there's a LOT of engineering behind ClearType, including the use of conceptual 'perfect' display which is down-transformed to match the actual display, and then reverse-transformed to allow tuning to match the conceptual display as much as possible (ie. with a minimum of signal loss). All heavy signal processing stuff.
Simon
Coming soon - pyrogyra
You mean, their ineptness in showing red, green, and blue combining to create black?
Win dain a lotica, en vai tu ri silota
Look at such screens from 2 ft long enough and soon you won't be able to discern the individual pixels without corrective optics!
Genetic condition? I doubt it. Unless something really went wrong in Singapore. Myopia seems almost endemic there. http://www.snec.com.sg/cec/childhood_myopia.htm
:(.
Myopia appears to be a result of too much "near work" - e.g. reading, watchmaking, embroidery. More environmental factors. Based on what I've observed that seems to be a far more plausible explanation.
There could be a genetic predisposition to _developing_ myopia after too much "near work". But "near work" seems to be the trigger. In fact I think it could be the other way round - e.g. some lucky people have genetic predisposition to not get myopia even after lots of "near work". So far scientists/doctors need to explore more on how a human eye develops/grows and maintains focus during that growth I believe they will find answers there - there are probably feedback loops and lots of close work could screw them up.
Personally I am not an eye doctor but I think that more kids are learning how to read at earlier ages, and their arms are just too short to put books at appropriate focal lengths comfortably! "Hold book 1.5-2 feet away" Yeah right, not easy when your arms aren't even that long, esp for precocious kids.
Someone should create special reading optics for those kids! Too late for me tho
Better color printers for computers are six colors now. Finally, you can get saturated reds.
None of this applies to light-emitting screens, which really are additive.
I don't like this layout. With the original triangular layout you could always find a triangle red Green and Blue. Using this layout you will not alway get blue next to green. I think this will show up in the final product.
At the next eco-hypocrisy-meeting, count the private jets used to get to the meeting. Should be interesting to see that